The proposal hypothesis is that even at low resolution, we can find and identify the correct structure model that combines good all-atom packing with relaxed local conformations and thus has the accuracy of a high- resolution structure solution. The problem to be solved is that biologically important large complexes of proteins and nucleic acids yield diffraction data only to low resolution (3 or worse), where functionally critical details of interatomic contacts cannot be seen directly and even overall folding patterns become uncertain. The diagnosis and correction methods we developed and applied successfully at higher resolution are extremely sensitive and can recognize a model with high-resolution validity and reject essentially any incorrect departure from that. However, it is a very difficult and currently unsolved search problem to identify such a model. The thrust of this proposal is therefore to develop and test search procedures for this task and determine the uniqueness of the solutions. These ideas contradict the standard paradigm by proposing that the crystallographically invisible hydrogen atoms can be effectively used even at low resolution, that only a small range of models are compatible both with the density and with known molecular properties, and therefore that detailed accuracy might actually be attainable even at low resolution.